Skip to main content
Glama
Browser Intelligence MCP

Browser Intelligence MCP

by JaDi03

Browser Intelligence MCP

A Model Context Protocol (MCP) server that gives AI agents real browser capabilities: take screenshots, scrape structured content, and run visual page tests.


Overview

Browser Intelligence MCP is a standard MCP server built with Node.js and TypeScript. It exposes three powerful tools that allow AI agents to interact with the real web through a headless Chromium browser powered by Playwright.

This server is the demo component of a per-usage billing infrastructure for agent-native services.


Related MCP server: MCP Browser Screenshot Server

Tools

Tool

Description

take_screenshot

Captures a full or viewport screenshot of any URL as a PNG image

scrape_page

Extracts structured content (title, description, links) from any URL

run_visual_test

Runs a set of checks against a page and returns pass/fail results


Requirements

  • Node.js >= 22

  • npm >= 10


Getting Started

# 1. Clone the repository
git clone https://github.com/your-org/browser-intelligence-mcp.git
cd browser-intelligence-mcp

# 2. Install dependencies
npm install

# 3. Install Playwright browsers
npx playwright install chromium

# 4. Copy environment variables
cp .env.example .env

# 5. Run in development mode
npm run dev

MCP Configuration

Add to your MCP client config (e.g., Claude Desktop claude_desktop_config.json):

{
  "mcpServers": {
    "browser-intelligence": {
      "command": "node",
      "args": ["dist/index.js"],
      "cwd": "/path/to/browser-intelligence-mcp"
    }
  }
}

Scripts

Command

Description

npm run dev

Run in development mode with hot reload

npm run build

Compile TypeScript to JavaScript

npm start

Run compiled production build

npm run lint

Run ESLint checks


License

MIT

Available Tools

3 tools
run_visual_testB

Runs a set of visual checks against a URL using a real browser. Supported check types: 'has_text' (text appears on page), 'has_element' (CSS selector exists), 'title_contains' (page title contains value), 'status_ok' (HTTP response is 2xx/3xx). Returns a detailed pass/fail report for each check.

Inputs: url (required), checks (required).

scrape_pageA

Navigates to a URL with a real browser and extracts structured content: page title, meta description, visible body text (first 5000 chars), and up to 30 links. Optionally provide a CSS selector to extract a specific element's text.

Inputs: url (required), selector.

take_screenshotA

Captures a PNG screenshot of any URL using a real headless browser. Returns the image as a base64-encoded string. Use fullPage=true to capture the entire scrollable page.

Inputs: url (required), fullPage.

Tool Schema Changelog

Recent tool additions, removals, and schema changes observed during successful MCP inspections.

Loading tool schema changelog…

    TDQS

    A4/5.0

    Scored across 3 tools

    Disambiguation5/5

    Each tool has a clear, non-overlapping purpose: take_screenshot captures visual state, scrape_page extracts content and links, and run_visual_test performs assertion-based checks. No ambiguity in usage.

    Naming Consistency5/5

    All three tools follow a consistent verb_noun pattern (take_screenshot, scrape_page, run_visual_test) with clear, descriptive verbs. Naming is uniform and predictable.

    Tool Count5/5

    With 3 tools, the set is minimal but well-scoped for a focused browser intelligence server. Each tool serves a distinct core function (capturing, extracting, testing) and earns its place without unnecessary bloat.

    Completeness4/5

    The surface covers the primary browser intelligence workflows (screenshot, scrape, and visual checks). Minor gaps exist—such as no support for full HTML extraction or waiting for dynamic content—but for the stated purpose, the coverage is solid and workable.

    Related MCP Connectors

    Related MCP Servers